CN111464442B - Method and device for routing data packet - Google Patents

Abstract

The disclosure relates to a method and a device for routing a data packet, and belongs to the technical field of network communication. The method comprises the following steps: when a data packet of an operation task type is received, determining a first operation task type corresponding to the data packet; determining at least one other node corresponding to the first operation task type and the operation performance corresponding to the at least one other node based on a first corresponding relation of the operation task type, the other nodes and the operation performance which is obtained in advance; determining a target node in at least one other node based on the operation performance corresponding to the at least one other node and the link state between the local node and the at least one other node; and determining the address of the target node as the destination address of the data packet, and forwarding the data packet based on the destination address. By adopting the method and the device, the destination node can be ensured to be capable of rapidly completing the operation task and feeding back the operation result to the initiating node of the data packet.

Description

Method and apparatus for routing data packets

technical field

The present disclosure relates to the technical field of network communication, in particular to a method and device for routing data packets.

Background technique

When the routing node receives any data packet, the routing node can determine the next-hop routing node according to the current network condition, and forward the data packet to the next-hop routing node. Among them, the data packet may carry information indicating the target data to be obtained from the data node, carrying instant messaging information, or carrying information indicating that the computing node is required to perform the computing task of the target type (the data packet carrying such information may be called data packets of the computing task type), etc.

There can be multiple computing nodes that can perform the same type of computing task. For a data packet of the computing task type, as long as it is forwarded to any computing node that can perform the corresponding computing task, the computing node can perform the computing task and output the calculation result. For example, the data packet of the image recognition type carries the image to be recognized. When the computing node receives the data packet of the image recognition type, it can obtain and recognize the image to be recognized, and return the recognition result to the initiating node of the task.

In the process of realizing the present disclosure, the inventors found at least the following problems:

In related technologies, in the process of routing data packets, how to forward data packets is determined entirely according to current network conditions. For example, computing nodes that can execute computing task A include computing node B and computing node C. According to the current network status, it is determined to forward the data packet corresponding to computing task A to computing node C corresponding to the path with the best current network status. However, in practical applications, the calculation node corresponding to the path with the best current network condition is not necessarily the optimal node for performing calculation tasks. If the above situation occurs, the originating node of the computing task type data packet will have to wait for a long time before obtaining the desired result.

Contents of the invention

In order to overcome the problems existing in related technologies, the present disclosure provides the following technical solutions:

In a first aspect, a method for routing data packets is provided, the method comprising:

When receiving a data packet of a computing task type, determining a first computing task type corresponding to the data packet;

Determine at least one other node corresponding to the first computing task type and computing performance corresponding to the at least one other node based on the pre-acquired first correspondence between the computing task type, other nodes, and computing performance;

Based on the computing performance corresponding to the at least one other node and the link status between the local node and the at least one other node, in the at least one other node, determine the target node;

The address of the target node is determined as the destination address of the data packet, and the data packet is forwarded based on the destination address.

Through the method provided by the embodiments of the present disclosure, in the process of routing data packets, in addition to considering the current network conditions, it is also based on the computing performance of each node that can perform the computing task indicated by the data packet of the computing task type, Determine the destination node. In this way, it can be ensured that the destination node can quickly complete the operation task and feed back the operation result to the originating node of the data packet, thereby shortening the waiting time of the originating node of the data packet.

In a possible implementation manner, the computing performance includes computing delay, and the link state includes data packet round-trip A link state between at least one other node in which a target node is determined, comprising:

For each other node, determine the sum of the operation delay corresponding to the other node and the round-trip delay of the data packet between the local node and the other node;

Among the at least one other node, it is determined that the node corresponding to the minimum sum value is the target node.

The local node determines the nodes capable of performing the computing task corresponding to each computing task type, and the computing delay required by each node to execute the computing task of each computing task type and the data packet round-trip delay between the local nodes. For each other node, determine the sum of the operation delay corresponding to the other node and the round-trip delay of the data packet between the local node and the other node, and determine the node corresponding to the minimum sum value as the target node among at least one other node.

In a possible implementation, the method further includes:

When the local node starts, for each other node, send a computing task type query request to the other node, receive at least one computing task type returned by the other node, and send the at least one computing task type to the other node A computing performance query request corresponding to the task type, receiving the computing performance corresponding to the at least one computing task type returned by the other nodes;

Based on at least one computing task type corresponding to each other node and computing performance corresponding to the at least one computing task type, a first correspondence between computing task types, other nodes, and computing performance is established.

Before the local node is started, the first corresponding relationship is not stored in the local node, and the first corresponding relationship needs to be established after the local node is started.

In a possible implementation manner, the other node and the local node belong to the same preset network area, the computing performance includes load information and computing delay, and the receiving the at least Computing performance corresponding to a computing task type, including:

receiving current load information corresponding to the at least one computing task type returned by the other nodes;

The method also includes:

According to the second corresponding relationship between the pre-stored load information and the operation delay, the operation delay corresponding to the current load information is determined as the operation delay corresponding to the at least one operation task type.

The historical load information and operation delay related data can be imported and stored in the local node in advance, and the local node can fit these historical data to determine the relationship between the load information and the operation delay, and then, when at least one When the current load information corresponding to the task type is calculated, the calculation delay corresponding to the current load information can be determined. Furthermore, the computing delay corresponding to at least one computing task type can be determined.

In a possible implementation manner, the other node and the local node do not belong to the same preset network area, the operation performance includes operation delay, and the receiving the at least one operation returned by the other node Computing performance corresponding to the task type, including:

receiving the computing delay corresponding to the at least one computing task type returned by the other node.

In a possible implementation manner, the number of updates of computing performance is also stored in the first correspondence, and the method further includes:

When receiving the computing performance query request sent by any other node among the other nodes, obtain the query computing task type and the corresponding update times carried in the computing performance query request, wherein the computing performance query request uses To instruct to query the computing performance of other nodes belonging to the same predetermined network area as the local node;

In the first correspondence, determine the number of updates of the computing performance corresponding to the query computing task type;

If the determined update times are greater than the update times carried in the computing performance query request, send the computing performance corresponding to the query computing task type and the determined update times to any other node.

When the local node is started, the first corresponding relationship can be initially established. However, since the operation delay is not fixed, but changes dynamically according to specific conditions as time goes by, the operation delay needs to be updated.

In a possible implementation, the method further includes:

Whenever receiving a target data packet whose destination address is another node belonging to the same preset network area as the local node, determine the second computing task type corresponding to the target data packet, and send a message to the Other nodes in the same preset network area forward the target data packet;

When the operation result corresponding to the target data packet returned by other nodes belonging to the same preset network area as the local node is received, determine the operation time between the time point of forwarding the target data packet and the current time point Delay, determining the operation delay as the operation performance corresponding to the second operation task type;

using the computing performance corresponding to the second computing task type to replace the computing performance corresponding to the second computing task type corresponding to other nodes in the first correspondence that belong to the same preset network area as the local node, And update the number of updates of the replaced computing performance in the first correspondence.

A second aspect provides a device for routing data packets, the device includes at least one module, and the at least one module is used to implement the method for routing data packets provided in the first aspect above.

In a third aspect, a node is provided, the node includes a processor and a memory, and the processor is configured to execute instructions stored in the memory; the processor implements the routing of data packets provided in the first aspect above by executing the instructions method.

In a fourth aspect, a computer-readable storage medium is provided, including an instruction, and when the computer-readable storage medium runs on a node, the node executes the method described in the above-mentioned first aspect.

In a fifth aspect, a computer program product containing instructions is provided, and when the computer program product is run on a node, the node is made to execute the method described in the first aspect above.

The technical solutions provided by the embodiments of the present disclosure may include the following beneficial effects:

Through the method provided by the embodiments of the present disclosure, in the process of routing data packets, in addition to considering the current network conditions, it is also based on the computing performance of each node that can perform the computing task indicated by the data packet of the computing task type, Determine the destination node. In this way, it can be ensured that the destination node can quickly complete the operation task and feed back the operation result to the originating node of the data packet, thereby shortening the waiting time of the originating node of the data packet.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description serve to explain the principles of the disclosure. In the attached picture:

Fig. 1 is a schematic flowchart of a method for routing data packets according to an exemplary embodiment;

Fig. 2 is a schematic flowchart of a method for routing data packets according to an exemplary embodiment;

Fig. 3 is a schematic diagram showing a network structure according to an exemplary embodiment;

Fig. 4 is a schematic flowchart of a method for routing data packets according to an exemplary embodiment;

Fig. 5 is a schematic flowchart of a method for routing data packets according to an exemplary embodiment;

Fig. 6 is a schematic structural diagram of a device for routing data packets according to an exemplary embodiment;

Fig. 7 is a schematic structural diagram of a node according to an exemplary embodiment.

By means of the above-mentioned drawings, certain embodiments of the present disclosure have been shown and will be described in more detail hereinafter. These drawings and written description are not intended to limit the scope of the disclosed concept in any way, but to illustrate the disclosed concept for those skilled in the art by referring to specific embodiments.

Detailed ways

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatuses and methods consistent with aspects of the present disclosure as recited in the appended claims.

An exemplary embodiment of the present disclosure provides a method for routing data packets, as shown in FIG. 1 , the processing flow of the method may include the following steps:

Step S110, when a data packet of a computing task type is received, determine a first computing task type corresponding to the data packet.

In implementation, when the local node receives the data packet, it can determine the task type corresponding to the received data packet, including indicating that it needs to obtain target data from the data node, transmit instant messaging information, or indicate that the computing node needs to perform the computing task of the target type (This kind of data packet may be called a computing task type data packet) and so on.

When the local node receives a data packet of a computing task type, it may determine a first computing task type corresponding to the data packet. In practical applications, when the local node receives the data packet, it can obtain the Internet Protocol Address (Internet Protocol Address, IP) carried in the packet header of the data packet. Next, the local node can determine the type of the carried IP address, and if the carried IP address is the IP address of any node, the data packet is forwarded based on the IP address of any node. If the carried IP address corresponds to any computing task, it can be determined that the received data packet is a data packet of the computing task type.

Since there can be many types of computing tasks, different computing tasks need to be distinguished through computing task type identifiers. The local node can obtain the computing task type identifier carried in the packet header of the data packet to determine the first computing task type corresponding to the data packet. It should be noted that in the local node, a new routing protocol needs to be run to obtain the computing task type identifier carried in the packet header of the data packet based on the new routing protocol, and based on the computing task type identifier, perform routing processing on the data packet.

Step S120, based on the pre-acquired first correspondence between computing task types, other nodes and computing performance, determine at least one other node corresponding to the first computing task type and computing performance corresponding to the at least one other node.

During implementation, new routing table items can be added, including computing task types and computing performance. A first correspondence between computing task types, other nodes, and computing performance may be pre-established in the local node, and based on the first correspondence, at least one other node corresponding to the first computing task type and computing performance corresponding to at least one other node are determined.

The computing tasks that can be performed by different computing nodes can be one type or multiple types, and the computing tasks that can be performed by different computing nodes can be the same or different. Therefore, it is first possible to determine which nodes can perform the computing task of the first computing task type, and then select the optimal node among these nodes capable of performing the computing task of the first computing task type.

For example, if the user wants the cloud to help identify all the people in the target image, this can be achieved by sending a data packet that identifies all the people in the target image. When the local node receives the data packet for identifying all the characters in the target image, it can obtain the identification of the computing task type in the data packet. Based on the computing task type identifier, find nodes that can execute the computing task corresponding to the computing task type identifier, including node A, node B, and node C. Then, computing performances corresponding to these nodes can be determined respectively. Wherein, the computing performance may include parameter information such as computing time delay, which can reflect the execution capabilities of different nodes in executing computing tasks.

Step S130, based on the computing performance corresponding to the at least one other node and the link status between the local node and the at least one other node, among the at least one other node, determine a target node.

In an implementation, the routing table entry may also include link states corresponding to different nodes, and the link state may include round-trip delays of data packets between the local node and other nodes. In practical applications, the local node may determine at least one other node corresponding to the first computing task type, and then determine link states between the local node and the at least one other node respectively. The target node is determined in the at least one other node based on factors such as the computing performance corresponding to the at least one other node and the link status between the local node and the at least one other node.

Optionally, the computing performance includes computing delay, and the link status includes data packet round-trip delay. Step S130 may include: for each other node, determining the computing delay corresponding to other nodes and the data between the local node and other nodes The sum value of the packet round-trip delay; among at least one other node, determine the node corresponding to the minimum sum value as the target node.

In implementation, as shown in Table 1, correspondences including computing task types, other nodes, computing delays, and data packet round-trip delays between the local node and other nodes may be established in advance.

Table 1

Based on Table 1, it is possible to determine the nodes capable of performing the computing tasks corresponding to each computing task type, as well as the computing delay required by each node to perform computing tasks of each computing task type and the round-trip delay of data packets between local nodes . For each other node, determine the sum of the operation delay corresponding to the other node and the round-trip delay of the data packet between the local node and the other node, and determine the node corresponding to the minimum sum value as the target node among at least one other node.

As shown in FIG. 2 , the node corresponding to each computing task type may be a computing node belonging to the same preset network area as the local node, or a routing node not belonging to the same preset network area as the local node. If the node corresponding to the computing task type is a routing node M that does not belong to the same preset network area as the local node, the data packet needs to be forwarded to the routing node M, and then the routing node M forwards the data packet to the routing node M that belongs to the routing node M. Computing nodes in the same preset network area.

Step S140, determining the address of the target node as the destination address of the data packet, and forwarding the data packet based on the destination address.

In implementation, after the target node is determined, the address of the target node may be queried, the address of the target node is determined as the destination address of the data packet, and the data packet is forwarded based on the destination address. When other routing nodes receive the data packet of the computing task type with the address of the target node as the destination address, they can forward the data packet only based on the network status, and finally forward the data packet to the destination node.

When the destination node receives a data packet of the computing task type with its own address as the destination address, it can directly forward the data packet to the local computing node, or re-determine the corresponding Whether the sum of the operation delay and the round-trip delay of the data packet is still the smallest, if not, re-determine the destination node.

The data packet is finally processed by the computing node, and the processing result is returned to the routing node that belongs to the same preset network area as the computing node, and then the routing node returns the processing result to the initiating node of the computing task according to the original path.

Before all routing nodes are started, area planning and level planning can be performed for these routing nodes. As shown in Figure 3, routing nodes can be laid out in a distributed and central manner. There can be a superior-subordinate relationship between routing nodes. The upper-level routing node is the central controller of the lower-level routing node. The lower-level routing node can accept the control of the upper-level routing node. Node information, to avoid lower-level routing nodes from probing the same-level routing nodes one by one to obtain node information, which can improve the efficiency of obtaining node information, and routing nodes that are mutually at the same level can exchange routing information with each other.

The upper and lower routing nodes can be laid out in a central structure, and the routing nodes at the same level can be laid out in a distributed structure. As the hierarchy becomes higher and higher, the number of routing nodes decreases, and the routing nodes converge as the hierarchy increases, and finally the entire network composed of routing nodes takes on a cone shape.

All the nodes implementing the new routing protocol in the above network can be used as the local nodes in the method provided by the embodiments of the present disclosure. Before the local node is started, the first corresponding relationship is not stored in the local node, and the first corresponding relationship needs to be established after the local node is started.

Optionally, the method provided by the embodiments of the present disclosure may further include: when the local node is started, for each other node, sending a computing task type query request to other nodes, receiving at least one computing task type returned by other nodes, and sending to other nodes A node sends a computing performance query request corresponding to at least one computing task type, and receives computing performance corresponding to at least one computing task type returned by other nodes; based on at least one computing task type corresponding to each other node and at least one computing task type corresponding to Computing performance, establish the first correspondence between computing task types, other nodes and computing performance.

In implementation, when the local node starts, the upper node of the local node can detect the start of the local node, and the upper node can send the node information of the peer nodes of the local node to the local node, so that the local node can determine the peer node. Among them, nodes at the same level include computing nodes that belong to the same preset network area as the local node, and routing nodes that do not belong to the same preset network area as the local node, such as node A, node B, node C, and node D in Table 1.

The local node can establish table 1 based on nodes at the same level. At this time, there are only other nodes in table 1, and the initial values of other table items are all 0. Local nodes can send computing task type query requests to other nodes, receive at least one computing task type returned by other nodes, send computing performance query requests corresponding to at least one computing task type to other nodes, and receive at least one computing task type returned by other nodes corresponding computing performance. Next, the local node may establish a first correspondence between computing task types, other nodes, and computing performance based on at least one computing task type corresponding to each other node and computing performance corresponding to the at least one computing task type.

For computing nodes that belong to the same preset network area as the local node, the computing performance may include load information and computing delay, and the step of receiving computing performance corresponding to at least one computing task type returned by other nodes may specifically include: receiving the computing performance returned by other nodes Current load information corresponding to at least one computing task type. Next, the local node may determine the computing latency corresponding to the current load information as the computing latency corresponding to at least one computing task type according to the second correspondence between the pre-stored load information and computing latency.

The historical load information and operation delay related data can be imported and stored in the local node in advance, and the local node can fit these historical data to determine the relationship between the load information and the operation delay, and then, when at least one When the current load information corresponding to the task type is calculated, the calculation delay corresponding to the current load information can be determined. Furthermore, the computing delay corresponding to at least one computing task type can be determined.

Computational delay can simply and directly reflect the ability of a computing node to perform a certain computing task. Although there are many factors that affect a computing node's execution of a certain computing task, they can all be directly reflected in the computing delay. The shorter the computing delay, the stronger the ability of the computing node to perform a certain computing task. Factors that may affect the execution of a computing task by a computing node include performance of a central processing unit (Central Processing Unit, CPU), performance of a graphics processing unit (Graphics Processing Unit, GPU), real-time load, and the like. In practical applications, some computing tasks have higher requirements on the performance of the CPU, but not high requirements on the performance of the GPU. Some computing tasks have high performance requirements on the GPU, but not high performance requirements on the CPU. For example, computing tasks of the image recognition type have higher requirements on the performance of the GPU.

For routing nodes that do not belong to the same preset network area as the local node, the computing performance may include computing delay, and the step of receiving computing performance corresponding to at least one computing task type returned by other nodes may include: receiving at least one computing returned by other nodes The computing latency corresponding to the task type.

The computing node N that belongs to the same preset network area as the routing node M can maintain the computing performance of the computing node N by the routing node M, while the routing node P that does not belong to the same preset network area as the routing node M, because the routing node P maintains Therefore, the routing node M can directly detect the computing performance of the computing node Q from the routing node P.

For the round-trip delay of data packets between the local node and other nodes, the round-trip delay of data packets between the local node and other nodes can be determined through interactive methods such as Internet packet explorer (Packet Internet Groper, PING) according to the preset period delay.

Through the above method, when the local node is started, the first corresponding relationship can be initially established. However, since the operation delay is not fixed, but changes dynamically according to the specific situation as time goes by, it is necessary to calculate the operation delay. renew.

For a computing node belonging to the same preset network area as the local node, whenever receiving a target data packet whose destination address is another node belonging to the same preset network area as the local node, determine the second computing task corresponding to the target data packet type, and forward the target data packet to other nodes belonging to the same preset network area as the local node; when receiving the operation result corresponding to the target data packet returned by other nodes belonging to the same preset network area as the local node, determine The computing delay between the time point of forwarding the target data packet and the current time point is determined as the computing performance corresponding to the second computing task type; the computing performance corresponding to the second computing task type is used to replace the first corresponding relationship Neutralize the computing performance corresponding to the second computing task type corresponding to other nodes in the same preset network area where the local node belongs, and update the number of updates of the computing performance after replacement in the first correspondence.

Since the local node needs to forward the data packet of the computing task type for the computing node belonging to the same preset network area, when the computing node executes the computing task corresponding to the data packet of the computing task type, it can reflect the current status of the computing node, and the local node can These conditions are counted, and the computing performance corresponding to the computing nodes belonging to the same preset network area is updated.

For example, when a local node forwards a data packet for performing image recognition to a computing node belonging to the same preset network area, the time point of forwarding can be recorded, and when the recognition result returned by the computing node is received, the time point of returning the recognition result can be determined The calculation delay between the forwarding time point and the forwarding time point can determine how long the current calculation node needs to perform the calculation task of image recognition. The update times of computing performance are also stored in the first correspondence, and the update times may be increased by 1 whenever the local node updates the computing delays of computing nodes belonging to the same preset network area. The initial value of the number of updates can be set to 0. As shown in Table 2, it is the corresponding relationship among computing task types, other nodes, computing delays, packet round-trip delays between the local node and other nodes, and update times.

Table 2

For routing nodes that do not belong to the same preset network area as the local node, the local node may send a detection packet (also called a computing performance query request) to the routing node to obtain computing performance of the node to be updated. The number of updates of computing performance is also stored in the first correspondence, and the method provided by the embodiment of the present disclosure may further include: when receiving a computing performance query request sent by any other node among other nodes, obtaining The carried query operation task type and the corresponding update times, wherein the operation performance query request is used to indicate the operation performance of other nodes belonging to the query and the local node in the same preset network area; in the first correspondence, the query operation task is determined The number of updates of the computing performance corresponding to the type; if the determined number of updates is greater than the number of updates carried in the computing performance query request, send the query computing performance corresponding to the computing task type and the determined number of updates to any other node.

As shown in Figure 4, if the current local node needs to update the computing capabilities of any other node among other nodes, it can first determine all computing task type identifiers and corresponding of updates. Carry the identification of the computing task type associated with any other node and the corresponding update times in the detection packet, and send it to any other node.

After any other node receives the detection packet sent by the local node, it determines the number of updates corresponding to the computing task type corresponding to the computing node belonging to the same preset network area as any other node. If the determined number of updates is greater than the number of updates carried in the detection packet, it is determined that the computing performance corresponding to the corresponding computing task type needs to be updated. Any other node carries all the determined computing performance corresponding to the types of computing tasks that need to be updated and the number of updates recorded in any other node in the detection return packet, and sends it to the local node. It should be noted that if any other node has a target computing task type that is not recorded in the first correspondence of the local node, it also needs to be sent to the local node, so that the local node adds a record about the target computing task type.

As shown in Table 3, it is the corresponding relationship among computing task types, other nodes, computing delays, packet round-trip delays between the local node and other nodes, and update times.

table 3

As shown in Figure 5, when a user equipment (User Equipment, UE) device such as UE1 initiates a data packet of computing task type 1, the data packet arrives at node 1 (local node), and node 1 acts as a management node to perform routing processing for the data packet . Node 1 looks up Table 3, and determines that the nodes capable of performing the computing task corresponding to computing task type 1 include node local, node 2, node 3, and node 4. Calculate the sum of the operation delay and the data packet round-trip delay of each node, and find that the sum corresponding to node 3 is the smallest, and node 1 can forward the data packet to node 3.

When UE2 initiates a data packet of computing task type 2, the data packet arrives at node 1, node 1 looks up table 3, and determines that the nodes capable of performing computing tasks corresponding to computing task type 2 include node local, node 2, and node 3. Calculate the sum of the operation delay of each node and the round-trip delay of the data packet, and find that the sum corresponding to the node local is the smallest, and node 1 can forward the data packet to the node local.

After a period of time, when UE1 initiates a data packet of computing task type 1 again, the data packet arrives at node 1. Since node 3 carries a large load at this time, the computing delay rises to 50ms, so the sum corresponding to node 2 this time If the value is the smallest, node 1 can forward the packet to node 2.

Through the method provided by the embodiments of the present disclosure, in the process of routing data packets, in addition to considering the current network conditions, it is also based on the computing performance of each node that can perform the computing task indicated by the data packet of the computing task type, Determine the destination node. In this way, it can be ensured that the destination node can quickly complete the calculation task and feed back the calculation result to the originating node of the data packet, thereby shortening the waiting time of the originating node of the data packet.

Another exemplary embodiment of the present disclosure provides an apparatus for routing data packets, as shown in FIG. 6, the apparatus includes:

The determining module 610 is configured to, when receiving a data packet of a computing task type, determine a first computing task type corresponding to the data packet; based on the pre-acquired first correspondence between the computing task type, other nodes, and computing performance, determine At least one other node corresponding to the first computing task type and the computing performance corresponding to the at least one other node; based on the computing performance corresponding to the at least one other node and the relationship between the local node and the at least one other node The link state, in the at least one other node, determines the target node, which can specifically implement the determination function in the above steps S110-130, and other implicit steps.

The sending module 620 is configured to determine the address of the target node as the destination address of the data packet, and forward the data packet based on the destination address, which can specifically implement the sending function in the above step S140, and other implicit steps.

Optionally, the computing performance includes computing delay, the link state includes data packet round-trip delay, and the determining module 610 is configured to:

For each other node, determine the sum of the operation delay corresponding to the other node and the round-trip delay of the data packet between the local node and the other node;

Among the at least one other node, it is determined that the node corresponding to the minimum sum value is the target node.

Optionally, the device also includes:

The receiving module is configured to, when the local node is started, for each other node, send a computing task type query request to the other node, receive at least one computing task type returned by the other node, and send to the other node The computing performance query request corresponding to the at least one computing task type, receiving the computing performance corresponding to the at least one computing task type returned by the other nodes;

An establishment module, configured to establish a first correspondence between an operation task type, other nodes, and operation performance based on at least one operation task type corresponding to each other node and the operation performance corresponding to the at least one operation task type.

Optionally, the other node and the local node belong to the same preset network area, the computing performance includes load information and computing delay, and the receiving module is configured to receive the at least The current load information corresponding to a computing task type;

The determination module 610 is further configured to determine the operation delay corresponding to the current load information as the operation time corresponding to the at least one operation task type according to the second correspondence between the pre-stored load information and the operation delay. delay.

Optionally, the other nodes and the local node do not belong to the same preset network area, the computing performance includes computing delay, and the receiving module is configured to:

receiving the computing delay corresponding to the at least one computing task type returned by the other node.

Optionally, the number of updates of computing performance is also stored in the first correspondence, and the device further includes:

An acquisition module, configured to acquire the query operation task type and the corresponding update times carried in the operation performance query request when receiving the operation performance query request sent by any other node in the other nodes, wherein the The computing performance query request is used to instruct to query the computing performance of other nodes belonging to the same preset network area as the local node;

The determining module 610 is further configured to determine, in the first correspondence, the number of updates of the computing performance corresponding to the query computing task type;

The sending module 620 is further configured to, when the determined update times are greater than the update times carried in the computing performance query request, send the computing performance corresponding to the query computing task type and the determined of updates.

Optionally, the determining module 610 is further configured to determine the first node corresponding to the target data packet whenever receiving a target data packet whose destination address is another node belonging to the same preset network area as the local node. Two calculation task types, and forward the target data packet to other nodes belonging to the same preset network area as the local node; when receiving all the returned nodes in the same preset network area as the local node When calculating the operation result corresponding to the target data packet, determine the operation delay between the time point of forwarding the target data packet and the current time point, and determine the operation delay as the operation performance corresponding to the second operation task type ;

The device also includes:

An update module, configured to use the computing performance corresponding to the second computing task type to replace the second computing task type corresponding to other nodes in the first corresponding relationship that belong to the same preset network area as the local node corresponding computing performance, and update the number of updates of the replaced computing performance in the first correspondence.

It should be noted that the above-mentioned determining module 610 and sending module 620 may be implemented by a processor, or implemented by a processor in cooperation with a memory and a transceiver.

Through the device provided by the embodiments of the present disclosure, in the process of routing data packets, in addition to considering the current network conditions, it is also based on the computing performance of each node that can perform the computing task indicated by the data packet of the computing task type, Determine the destination node. In this way, it can be ensured that the destination node can quickly complete the operation task and feed back the operation result to the originating node of the data packet, thereby shortening the waiting time of the originating node of the data packet.

It should be noted that: when the device for routing data packets provided by the above-mentioned embodiments is used for routing data packets, the division of the above-mentioned functional modules is used as an example for illustration. In practical applications, the above-mentioned functions can be allocated by Completion of different functional modules means that the internal structure of the node is divided into different functional modules to complete all or part of the functions described above. In addition, the device for routing data packets provided by the above embodiment and the method embodiment for routing data packets belong to the same idea, and the specific implementation process thereof is detailed in the method embodiment, and will not be repeated here.

The node 700 may include a processor 710, a memory 740, and a transceiver 730, and the transceiver 730 may be connected to the processor 710, as shown in FIG. 7 . The transceiver 730 may include a receiver and a transmitter, and may be used to receive or send messages or data, and the transceiver 730 may be a network card. The node 700 may also include an acceleration component (which may be called an accelerator). When the acceleration component is a network acceleration component, the acceleration component may be a network card. The processor 710 may be the control center of the node 700, and uses various interfaces and lines to connect various parts of the entire node 700, such as the transceiver 730 and the like. In the embodiment of the present disclosure, the processor 710 may be a central processing unit (Central Processing Unit, CPU). Optionally, the processor 710 may include one or more processing units. The processor 710 may also be a digital signal processor, an application specific integrated circuit, a field programmable gate array, or other programmable logic devices. The node 700 can also include a memory 740, which can be used to store software programs and modules. The processor 710 executes various functional applications and data processing of the node 700 by reading the software codes and modules stored in the memory.

Other embodiments of the disclosure will be readily apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, and these modifications, uses or adaptations follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not disclosed in the present disclosure . The specification and examples are to be considered exemplary only, with the true scope and spirit of the disclosure indicated by the appended claims.

It should be understood that the present disclosure is not limited to the precise constructions which have been described above and shown in the drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims (22)

1. A method of routing packets, the method comprising:

when a data packet of an operation task type is received, determining a first operation task type corresponding to the data packet;

determining at least one other node corresponding to a first operation task type and operation performance corresponding to the at least one other node based on a first corresponding relationship of the operation task type, the other nodes and the operation performance, wherein under the condition that the at least one other node and the local node belong to the same preset network area, the operation performance comprises load information and operation delay, the operation delay of the first operation task type corresponding to the at least one other node is in a second corresponding relationship, the operation delay corresponding to the load information of the at least one other node is in the second corresponding relationship, and the second corresponding relationship is the corresponding relationship between the load information and the operation delay;

determining a target node in the at least one other node based on the operation performance corresponding to the at least one other node and the link states between the local node and the at least one other node respectively;

and determining the address of the target node as the destination address of the data packet, and forwarding the data packet based on the destination address.

2. The method of claim 1, wherein the computation performance comprises computation latency, and the link status comprises packet round trip latency, and wherein determining, in the at least one other node, a target node based on the computation performance corresponding to the at least one other node and the link status between the local node and the at least one other node comprises:

for each other node, determining the sum of the operation delay corresponding to the other node and the round-trip delay of the data packet between the local node and the other node;

and determining the node corresponding to the minimum sum value as a target node in the at least one other node.

3. The method of claim 1, further comprising:

when the local node is started, for each other node, sending an operation task type query request to the other node, receiving at least one operation task type returned by the other node, sending an operation performance query request corresponding to the at least one operation task type to the other node, and receiving operation performance corresponding to the at least one operation task type returned by the other node;

and establishing a first corresponding relation among the operation task type, the other nodes and the operation performance based on at least one operation task type corresponding to each other node and the operation performance corresponding to the at least one operation task type.

4. The method according to claim 3, wherein the other node and the local node belong to the same preset network area, the computation performance includes load information and computation delay, and the receiving the computation performance corresponding to the at least one computation task type returned by the other node includes:

receiving current load information corresponding to the at least one operation task type returned by the other nodes;

the method further comprises the following steps:

and determining the operation delay corresponding to the current load information as the operation delay corresponding to the at least one operation task type according to a second corresponding relation between pre-stored load information and operation delay.

5. The method according to claim 3, wherein the other node and the local node do not belong to the same preset network area, the computation performance includes computation delay, and the receiving the computation performance corresponding to the at least one computation task type returned by the other node includes:

and receiving the operation time delay corresponding to the at least one operation task type returned by the other nodes.

6. The method according to claim 1, wherein the first correspondence further stores the number of times of updating of the arithmetic performance, and the method further comprises:

when receiving an operation performance query request sent by any other node in the other nodes, acquiring a query operation task type and corresponding update times carried in the operation performance query request, wherein the operation performance query request is used for indicating the operation performance of other nodes in the same preset network area as the query and the local node;

in the first corresponding relation, determining the updating times of the operation performance corresponding to the query operation task type;

and if the determined updating times are larger than the updating times carried in the operation performance query request, sending the operation performance corresponding to the query operation task type and the determined updating times to any other node.

7. The method of claim 6, further comprising:

when receiving a target data packet with a destination address of other nodes belonging to the same preset network area as the local node, determining a second operation task type corresponding to the target data packet, and forwarding the target data packet to other nodes belonging to the same preset network area as the local node;

when receiving an operation result corresponding to the target data packet returned by other nodes belonging to the same preset network area as the local node, determining an operation delay between a time point of forwarding the target data packet and a current time point, and determining the operation delay as an operation performance corresponding to the second operation task type;

and replacing the operation performance corresponding to the second operation task type in the first corresponding relation and other nodes of which the local nodes belong to the same preset network area by using the operation performance corresponding to the second operation task type, and updating the number of times of updating the operation performance replaced in the first corresponding relation.

8. An apparatus for routing packets, the apparatus comprising:

the device comprises a determining module, a judging module and a judging module, wherein the determining module is used for determining a first operation task type corresponding to a data packet when the data packet of the operation task type is received; determining at least one other node corresponding to a first operation task type and operation performance corresponding to the at least one other node based on a first corresponding relationship of a pre-obtained operation task type, other nodes and operation performance, wherein under the condition that the at least one other node and the local node belong to the same preset network area, the operation performance comprises load information and operation delay, the operation delay of the first operation task type corresponding to the at least one other node is in a second corresponding relationship, the operation delay corresponding to the load information of the at least one other node is in the second corresponding relationship, and the second corresponding relationship is the corresponding relationship between the load information and the operation delay; determining a target node in the at least one other node based on the operation performance corresponding to the at least one other node and the link states between the local node and the at least one other node respectively;

and the sending module is used for determining the address of the target node as the destination address of the data packet and forwarding the data packet based on the destination address.

9. The apparatus of claim 8, wherein the computation performance comprises computation latency, wherein the link state comprises packet round trip latency, and wherein the determining module is configured to:

for each other node, determining the sum of the operation delay corresponding to the other node and the round-trip delay of the data packet between the local node and the other node;

and determining the node corresponding to the minimum sum value as a target node in the at least one other node.

10. The apparatus of claim 8, further comprising:

a receiving module, configured to send, for each other node, an operation task type query request to the other node, receive at least one operation task type returned by the other node, send an operation performance query request corresponding to the at least one operation task type to the other node, and receive operation performance corresponding to the at least one operation task type returned by the other node when the local node is started;

the establishing module is used for establishing a first corresponding relation among the operation task type, the other nodes and the operation performance based on at least one operation task type corresponding to each other node and the operation performance corresponding to the at least one operation task type.

11. The apparatus according to claim 10, wherein the other nodes and the local node belong to a same preset network area, the computation performance includes load information and computation delay, and the receiving module is configured to receive current load information corresponding to the at least one computation task type returned by the other nodes;

the determining module is further configured to determine, according to a second correspondence between pre-stored load information and operation delay, the operation delay corresponding to the current load information, as the operation delay corresponding to the at least one operation task type.

12. The apparatus of claim 10, wherein the other node and the local node do not belong to a same predetermined network area, the computation performance includes computation latency, and the receiving module is configured to:

and receiving the operation time delay corresponding to the at least one operation task type returned by the other nodes.

13. The apparatus according to claim 8, wherein the first correspondence relationship further stores the number of times of updating of the arithmetic performance, the apparatus further comprising:

the acquisition module is used for acquiring the query operation task type and the corresponding updating times carried in the operation performance query request when receiving the operation performance query request sent by any other node in the other nodes, wherein the operation performance query request is used for indicating the query and the operation performance of other nodes in the same preset network area of the local node;

the determining module is further configured to determine, in the first corresponding relationship, the number of times of updating the operation performance corresponding to the query operation task type;

the sending module is further configured to send the computation performance corresponding to the query computation task type and the determined update times to any other node when the determined update times are greater than the update times carried in the computation performance query request.

14. The apparatus according to claim 13, wherein the determining module is further configured to, each time a destination data packet is received, where the destination address of the destination data packet is a destination data packet of another node belonging to the same preset network area as the local node, determine a second operation task type corresponding to the destination data packet, and forward the destination data packet to another node belonging to the same preset network area as the local node; when receiving an operation result corresponding to the target data packet returned by other nodes belonging to the same preset network area as the local node, determining an operation delay between a time point of forwarding the target data packet and a current time point, and determining the operation delay as an operation performance corresponding to the second operation task type;

the device further comprises:

and the updating module is used for replacing the operation performance corresponding to the second operation task type in the first corresponding relation and other nodes of the local node belonging to the same preset network area by using the operation performance corresponding to the second operation task type, and updating the updating times of the replaced operation performance in the first corresponding relation.

15. A node, characterized in that the node comprises a processor, a memory and a transceiver, wherein:

the processor is used for determining a first operation task type corresponding to a data packet when the transceiver is controlled to receive the data packet of the operation task type; determining at least one other node corresponding to the first operation task type and operation performance corresponding to the at least one other node based on a first corresponding relationship of operation task types, other nodes and operation performance obtained in advance in the memory, wherein under the condition that the at least one other node and the local node belong to the same preset network region, the operation performance comprises load information and operation delay, the operation delay of the first operation task type corresponding to the at least one other node is in a second corresponding relationship, the operation delay corresponding to the load information of the at least one other node is in the second corresponding relationship, and the second corresponding relationship is the corresponding relationship between the load information and the operation delay; determining a target node in the at least one other node based on the operation performance corresponding to the at least one other node and the link states between the local node and the at least one other node respectively;

and the transceiver is used for determining the address of the target node as the destination address of the data packet and forwarding the data packet based on the destination address.

16. The node of claim 15, wherein the operational performance comprises an operational latency, wherein the link state comprises a packet round trip latency, and wherein the processor is configured to:

for each other node, determining the sum of the operation delay corresponding to the other node and the round-trip delay of the data packet between the local node and the other node;

and determining the node corresponding to the minimum sum value as a target node in the at least one other node.

17. The node according to claim 15, wherein the transceiver is configured to, when the local node is started, send, for each other node, an operation task type query request to the other node, receive at least one operation task type returned by the other node, send, to the other node, an operation performance query request corresponding to the at least one operation task type, and receive, from the other node, operation performance corresponding to the at least one operation task type;

the processor is used for establishing a first corresponding relation among the operation task type, the other nodes and the operation performance based on at least one operation task type corresponding to each other node and the operation performance corresponding to the at least one operation task type.

18. The node according to claim 17, wherein the other node and the local node belong to a same preset network area, the computation performance includes load information and computation delay, and the transceiver is configured to receive current load information corresponding to the at least one computation task type returned by the other node;

and the processor is configured to determine, according to a second correspondence between the load information and the operation delay pre-stored in the memory, the operation delay corresponding to the current load information as the operation delay corresponding to the at least one operation task type.

19. The node of claim 17, wherein the other node and the local node do not belong to the same predetermined network area, wherein the operational performance comprises an operational latency, and wherein the transceiver is configured to:

and receiving the operation time delay corresponding to the at least one operation task type returned by the other nodes.

20. The node according to claim 15, wherein the first mapping relationship further stores the number of times of updating the computation performance, and the processor is configured to, when receiving a computation performance query request sent by any one of the other nodes, obtain a query computation task type and a corresponding number of times of updating carried in the computation performance query request, where the computation performance query request is used to indicate the computation performance of other nodes in a same preset network area as the query and the local node; determining the updating times of the operation performance corresponding to the query operation task type in the first corresponding relation;

and the transceiver is used for sending the operation performance corresponding to the query operation task type and the determined updating times to any other node when the determined updating times are larger than the updating times carried in the operation performance query request.

21. The node of claim 20, wherein the processor is further configured to:

when receiving a target data packet with a destination address of other nodes belonging to the same preset network area as the local node, determining a second operation task type corresponding to the target data packet, and forwarding the target data packet to other nodes belonging to the same preset network area as the local node;

when receiving an operation result corresponding to the target data packet returned by other nodes belonging to the same preset network area as the local node, determining an operation delay between a time point of forwarding the target data packet and a current time point, and determining the operation delay as an operation performance corresponding to the second operation task type;

and replacing the operation performance corresponding to the second operation task type in the first corresponding relation and other nodes of which the local nodes belong to the same preset network area by using the operation performance corresponding to the second operation task type, and updating the number of times of updating the operation performance replaced in the first corresponding relation.

22. A computer-readable storage medium comprising instructions that, when executed on a node, cause the node to perform the method of any of claims 1-7.

Images